Thanks Riccardo,
That sounds good. Then I can have one global graph and my web frontend
(servlet) can be simplified and take advantage of the connection pool.
One question on updating this global graph.
To simplify things we are re-creating each customer's graph from scratch
daily
(OrientGraphNoTx ... declareIntent(new OMassiveInsert())
instead of detecting & applying differences. The new graph name is something
like custx_<uuid>. When finished we update some reference table that maps
each customer to their current graph, then drop the old graph.
In the proposed scenario, with a global graph and different class per
customer,
would I create new class each time I import new data for customerX, then
delete
the records with the old class, and rename the new class?
for example:
If my global graph was called cust_data and my Vertex classes were called
Cust1Page, Cust2Page,...
When import new data for Cust1 I would
1. create class Cust1Page_new extends V
2. import Cust1 data creating vertices of type Cust1Page_new
3. delete from cluster:Cust1Page;
4. alter class Cust1Page_new Cust1Page
Alan
On Wednesday, April 16, 2014 12:10:47 AM UTC-7, Riccardo Tasso wrote:
>
> Probably a good way to model your scenario is to subclass the V class.
> You can create one schema for the V class (author, lastmod, size,
> properties, ...) and then one class for each customer (Customer1 extends
> V, Customer2 extends V, ..., Customer3000 extends V).
>
> Advantages:
>
> - every subclass will inherit the parent schema
> - it should be faster iterate over all the edges of only one subclass
> ("SELECT FROM CustomerN ..."), or in general perform any kind of operation
> shoul be less complex
> - you can work at global level, if you want ("SELECT FROM V ...")
>
> I know that this seems not the answer to your question. Anyway I think
> that you will have benefits with a multi-class model, especially if each
> customer will access more frequently his own partion.
>
> Cheers,
> Riccardo
>
>
> 2014-04-15 23:54 GMT+02:00 Alan Miller <[email protected] <javascript:>>
> :
>
>> Hi,
>>
>> Is it possible to use a connection pool with a partitioned graph?
>>
>> Consider this:
>> Let's say I wanted to model the contents of 3000 websites in a single
>> graph.
>> The Vertex class in the graph might be called: "Page" with author,
>> lastmod, size, properties.
>> The Edge class might be called "LINKSTO" (e.g. 0:many (Page -- LINKSTO
>> --> Page) relationship.
>>
>> Since each website is a separate customer, I wanted to avoid having 3000
>> graphs
>> and use a Partitioned-Graph.
>> https://github.com/orientechnologies/orientdb/wiki/Partitioned-Graphs
>> So I'd have one graph for all 3000 customers but 3000 customerid (or
>> graph user).
>>
>> Although I anticipate my site to be low traffic (at any given time < 100
>> customers)
>> how can I effectively use the connection pool?
>>
>> The connection pool works at the DB level right? Do I need a separate
>> pool for every customer?
>>
>> Per the wiki
>>
>> database = new ODatabaseDocumentTx("remote:localhost/demo");
>> database.setProperty("minPool", 2);
>> database.setProperty("maxPool", 5);
>>
>> database.open("admin", "admin");
>>
>> Alan
>>
>> --
>>
>> ---
>> You received this message because you are subscribed to the Google Groups
>> "OrientDB" group.
>> To unsubscribe from this group and stop receiving emails from it, send an
>> email to [email protected] <javascript:>.
>> For more options, visit https://groups.google.com/d/optout.
>>
>
>
--
---
You received this message because you are subscribed to the Google Groups
"OrientDB" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
For more options, visit https://groups.google.com/d/optout.
